$\rm R\sb2Fe\sb{17}N\sb{y}$ and $\rm R\sb2Fe\sb{17}C\sb{y}$ have attracted much attention because of their excellent intrinsic magnetic properties. However, these nitrides and carbides prepared by a gas-phase reaction have a major drawback, namely their chemical or structural instability at high temperature. We have discovered that Si substitution is able to stabilize the 2:17 structure of the carbides. A new series, $\rm R\sb2Fe\sb{15}Si\sb2C\sb{y}$ (R = Nd, Sm, Gd and Er), has been prepared by heating powders of their parents in CH$\sb4$ gas at 700$\sp\circ$C instead of usual 500$\sp\circ$C. For these carbides, the carbon concentrations are close to the theoretical value of y = 3 and the 2:17 structure is still retained up to 700$\sp\circ$C. Therefore, it is possible to make these carbides into the anisotropic permanent magnets by aligning and then sintering. The crystal structures, the Curie temperatures, the saturation magnetizations, the temperature dependences of the magnetization, the magnetic anisotropies and the spin reorientation transitions of $\rm \sb2Fe\sb{15}Si\sb2C\sb{y}$ have been studied by x-ray diffraction and magnetic measurements. These carbides have elevated Curie temperatures as well as other magnetic properties that are modified as compared to their parents. For $\rm Sm\sb2Fe\sb{16}SiC\sb{2.7},$ the Curie temperature is 660 K; the anisotropy field and saturation magnetization at room temperature are 95 kOe and 92 emu/g, respectively. The origin of the unexpected decrease in the magnetization is discussed. The hyperfine fields, the distribution of Fe and Si atoms on each Fe site and the temperature dependence of the hyperfine parameters have been studied by Mossbauer spectra in an attempt to unravel the magnetic properties of these carbides on an atomic scale. The exchange integrals as a function of the distance between the Fe-Fe pairs have been obtained; the relationships between the Curie temperature and the site occupancy of Si atoms have been found. A negative exchange interaction model is proposed to interpret the origin in the greatly elevated Curie temperature for the nitrides and carbides and to predict the Curie temperatures of $\rm R\sb2(FeSi)\sb{17}$ compounds; this model is consistent with the experimental results.